DRAFT WORKING PAPER ONLY
Tying the Dictator’s Hands: Leadership Survival in
Authoritarian Regimes
By Erica Frantz
Department of Political Science, UCLA
[email protected]
Electronic copy of this paper is available at: http://ssrn.com/abstract=975161
Abstract: In this paper, I examine one of the most important elements of authoritarian
politics: leadership survival. I argue that a key actor in understanding leadership survival
in dictatorships is the elite coalition—the group of individuals a dictator relies on for
support. Whether elites are bound together by a dominant institution, such as a party or
military, affects their ability to overthrow the dictator. I look at elite coalitions in
personalist, single-party and military dictatorships. Due to institutional differences in the
nature of the elite coalition, elites in military dictatorships should have the greatest ability
to oust the dictator, followed by elites in single-party dictatorships, and lastly elites in
personalist dictatorships. I test this expectation using a survival model and find
substantial support for my argument. Military dictators face the highest risk of removal
from office and personalist dictators face the lowest.
2
Electronic copy of this paper is available at: http://ssrn.com/abstract=975161
In comparison to democratic systems, we know very little about how dictatorships
work, who the key political actors are, and where the locus of power rests. In fact,
authoritarian rule is one of the least-studied areas of political science.1 Yet, authoritarian
government has been the norm for most of history. As late as the 1970’s, autocracy was
more common than democracy. At the turn of the century, even by a modest estimate,
nearly a quarter of countries were governed under authoritarian rule.2 The Chinese
communist regime alone rules nearly a quarter of the world’s population (Brooker 2000,
1). Given that dictatorships are so widespread, it is essential that we broaden our
understanding of the mechanics of authoritarian regimes.
In this paper, I examine one of the most important elements of authoritarian
politics: leadership survival. I argue that under dictatorships, one of the key political
actors is the elite coalition—the group that a dictator relies on for support. Indeed, the
vast majority of dictators are overthrown by their own inner circle (Svolik 2006).
Developing an understanding of a dictator’s supporting coalition is central to
understanding the sources of a dictator’s survival in power. Elites are not equally capable
of deposing dictators, however. The purpose of this study is to examine the conditions
under which it is easier for erstwhile support coalitions to topple authoritarian leaders—
as happened in the last decade to Liamine Zeroual in Algeria and Ibrahim Bare
Mainassara in Niger. How do elite coalitions differ across dictatorships and what do
these differences imply for the survival of the dictator?
This study is informed by the case study literature on dictatorships and by the
literature on particular authoritarian institutional structures. I combine the two to build a
As Przeworski noted in 2003, “Dictatorships are by far the most understudied area in comparative politics.
We need to start thinking about it”
(http://www.nyu.edu/gsas/dept/politics/faculty/przeworski/przeworski_munck.pdf).
2
This estimate was tabulated by using data from the World Bank (2003) and Geddes (2003).
1
3
framework for understanding the consequences of authoritarian institutions for the
survival of the dictator. The case study literature devoted to authoritarian regimes is
abundant. A multitude of scholars have examined personalist regimes, including
Hartlyn’s work on Rafael Trujillo of the Dominican Republic (1998) and Farouk-Slugett
and Slugett’s work on Saddam Hussein of Iraq (1987). Stepan (1974), Arceneaux (2001),
and Pinckney (1972) have provided excellent accounts of military dictatorships. They
use examples of Brazil, Argentina, and Ghana to detail the structure of military rule.
Cases devoted to single-party regimes are also widespread. See, for example, Fitzpatrick
(1982) for work on the Soviet Union and Zolberg (1966) for studies on West African
one-party states. Case studies such as these provide valuable insights into how
individual dictatorships function.
There is also an extensive body of work devoted to particular types of
authoritarian institutional structures. Bienen (1978), Clapham and Philip (1985), and
Nordlinger (1977), among many others, offer interesting theoretical analyses of military
rule. Similarly, Magaloni (2006), Huntington (1970), and Brooker (1995) provide
excellent insights into the behavior of single-party dictatorships. For work on less
institutionalized regimes, Chebabi and Linz (1998) offer a detailed study of personalist
rule, as do Bratton and Van de Walle (1997).
Though both literatures are independently quite rich, only recently have scholars
begun to integrate these studies and derive and test hypotheses of authoritarian politics
based on the institutional differences among these regimes. Past empirical work on
dictatorships has either left out the role of institutions or focused solely on one particular
type of dictatorship. There are, of course, notable exceptions. Bueno de Mesquita et al.’s
selectorate theory (2003) identifies two institutions of governance, the selectorate and the
4
winning coalition, that expose generic differences between democracy, monarchy,
military junta, and other forms of government. They claim that differences in the makeup of the winning coalition and the selectorate explain a variety of outcomes, such as
economic performance, political survival, and conflict behavior.3 Haber (2006) develops
a theoretical framework for understanding authoritarian politics. He argues that the
strategies played by dictators to stay in power lead to three different institutional
arrangements in authoritarian regimes: dictators who terrorize the leadership of their
launching organizations, dictators who co-opt the leadership of their launching
organization, and dictators who create a set of rival or complementary organizations.
Haber examines what these scenarios imply for political repression, property rights,
economic growth, and democratic transitions. Geddes (2003) shows that the institutional
structure of dictatorships helps to explain variations in their longevity. Single-party
regimes tend to be long-lasting and military regimes short lived.4
Such dictatorships
also experience different transitions to democracy (see also Haggard and Kaufman 1995).
Gandhi (2003) highlights the role of legislative institutions in determining economic
outcomes among dictatorships. She finds that economic growth is higher in
institutionalized dictatorships, particularly those that allow semi-autonomous parties
within a legislature.
In this study, I contribute to this recent line of work and apply an empirical
institutionalist perspective to the study of authoritarianism. My argument builds on the
well-established political science literature about domestic political institutions, which I
extend to authoritarian settings. I examine how the presence of institutional structures,
For operational difficulties applying the theory to dictatorships, see
http://personal.lse.ac.uk/kudamats/dew/dictatorship.htm
4
For an interesting analysis of the circumstances under which hegemonic parties do rupture, see Langston
2004.
3
5
like an established party or military, contrasts with less institutionalized authoritarian
environments, like that typically seen in personalist regimes. Whether elites are bound
together by a dominant institution affects their ability to overthrow the dictator. I argue
that it should be easier for elites in military dictatorships to overthrow the leader,
followed by elites in single-party dictatorships, and lastly elites in personalist
dictatorships. I test my argument quantitatively and find positive support: military
dictators face the highest risk of removal and personalist dictators face the lowest. These
findings have important implications for the study of authoritarian regimes. The greater
the ability of the elite coalition to unseat the dictator, the greater the bargaining power of
the elite coalition vis-à-vis the dictator.
One of the key insights in this paper is that leadership survival is often distinct
from regime survival. Authoritarian regimes frequently last well beyond the ouster of
any individual leader. I focus on the group that leaders rely on for support, rather than
the group that regimes rely on for support. These two groups are not always synonymous
in dictatorships. Failing to distinguish between the two can lead to misunderstandings of
both the causes of regime change and of leadership change.
The paper takes the following form. In the first section, I provide a review of the
relevant literature regarding leadership survival and elite coalitions in dictatorships. In
the second section, I put forth my theoretical argument for understanding the political
survival of dictators. In the third section, I test this argument quantitatively. And in the
last section, I conclude.
I. Leadership Survival and Elite Coalitions in Dictatorships.
What factors affect the survival of dictators? Prior studies of political survival in
dictatorships have emphasized the various strategies that dictators employ to stay in
6
power (Friedrich and Brzezinski 1965; Arendt 1951; Tullock 1987). Dictators are
characterized as stationary bandits (Olson 2000), maintaining their rule by using violence
to extract taxes. The prospect of popular uprising is highlighted as the central threat to
the dictator’s power (Acemoglu and Robinson 2001; Boix 2003; Sanhueza 1999). To
deter an uprising, dictators must repress some part of the population, while nurturing the
loyalty of others (Wintrobe 1998). As the time horizon for the dictator increases, the
quality of public good provision and economic policy will be positively correlated with
the payoffs of staying in power (McGuire and Olson 1996; Grossman and Noh 1994;
Galetovic and Sanhueza 1995).
Building on the scholarship on regime transitions (see Higley and Burton 1989,
Kugler and Feng 1999, and Haggard and Kaufman 1995), recent studies have recognized
the important role of the dictator’s elite supporters. Though this research is quite varied,
the identification of the authoritarian elite as a key political actor is central. Svolik
(2006), for example, models bargaining between the dictator and the ruling elite and
argues that, as a result of strategic behavior, two power-sharing regimes emerge:
contested and established dictatorships. In contrast, Gallego and Pitchik (2004) focus on
the role of the economic elite in determining the tenure of the dictator. They argue that in
order to survive leaders must offer the level of investment that equates the benefit and the
cost of a coup. In a different vein, Bertocchi and Spagat (2001) examine how power
structures affect governmental needs to transfer income. They posit that the more
powerful the ruling elite are, the more citizens are willing to accept co-optation and lower
levels of government transfers.
My method is novel in that I look at how the institutional structure of
authoritarian regimes influences the ability of elites to remove a dictator from power. I
7
focus on differences in the nature of elite coalitions in authoritarian regimes and the
implications of these differences for the survival of the dictator.
II. Theoretical Argument.
Dictatorships are often described as one-man rule, though all political leaders
require the support of some individuals in order to maintain their command. Though the
nature and method of selection of their supporting group may vary across regime types,
no leader rules entirely alone. In authoritarian regimes, the group of individuals whose
support the leader requires to stay in power is the elite coalition.5 Individuals agree to
support the leader in return for benefits. I look at how elite coalitions differ among the
following types of authoritarian regimes: personalist, military, and single-party.6 In
personalist regimes the elite coalition is the personalist clique, in military regimes it is the
military junta, and in single-party regimes it is the party cadre. The nature of the elite
coalition affects whether the coalition can depose the dictator.
There are two factors that affect the elite coalition’s ability to overthrow the
leader. The first is whether the coalition bargains with the dictator individually or as a
collective. I draw on the work of Geddes (2004) and argue that the behavior of elites in
military and single-party dictatorships is analogous to that of a union. Elites in military
and single-party regimes are bound together by an institution, the military and party
respectively, that unites them. This makes coordination among members of the coalition
less difficult. In addition, just as employers facing a unionized labor market have less
autonomy in hiring, leaders in military and single-party regimes have less control over
Though leaders in authoritarian regimes need the support of a certain number of individuals in order to
stay in power, they do not need the backing of each and every individual in the coalition. Most elites
support the dictator, but there may be some who do not. The exact number of elites required to keep the
dictator in power is unknown and varies from regime to regime. For in depth analyses of the process of
coup d'états, see Chorley (1973), Luttwak (1969), and Decalo (1976), to name a few.
6
I acknowledge that, in reality, leaders try and sometimes succeed in changing regime type. My argument
applies to bargaining in stable authoritarian regimes. For an analysis of the process of regime consolidation
in dictatorships and why some regimes personalize and others do not, see Geddes 2004.
5
8
membership in the elite coalition. They may try to influence membership in the coalition
via promotions in the military or within the party, but they do not necessarily have
absolute power over this. Leaders in military and single-party regimes have few other
pools from which to draw elites bar the military or party. They may try to create new
pools by forming a new party or paramilitary force, but these efforts are not always
successful. Leaders also must compete with other potential leaders to secure the
“employment” of the elites.7 The union behavior in military and single-party regimes has
two primary consequences: 1) elites face a lower coordination dilemma when it comes to
potentially ousting the leader, and 2) the leader has less control over the selection of
individuals who will comprise the coalition.
In contrast, in personalist dictatorships there is no institution that serves to unite
elites. Because of this, elites must overcome a greater coordination dilemma in order to
unseat the leader. Personalist leaders also have more control over the selection of
individuals who will comprise the coalition because they do not face a unionized type of
labor market. Instead, individuals must compete with one another to secure their spot in
the coalition. Personalist leaders often select to the coalition low skilled individuals who
are less likely to successfully unseat them. They also ensure that no individuals get too
powerful through divide and conquer strategies and frequent purges.8 For these reasons,
the probability the elite coalition can overthrow the leader tends to be greater in military
and single-party regimes than it is in personalist dictatorships.
The second factor that affects the ability of the elite coalition to oust the dictator
is the extent to which members of the elite coalition have control over the security forces.
For this reason, the leader’s situation is not monopsonistic.
Leaders in military and single-party regimes also engage in these strategies, though they tend to be less
successful.
7
8
9
Most coups are executed by military forces (Kebschull 1994).9 The more direct control
elites have over the security apparatus, the more the leader’s position is at risk. Elites in
military regimes tend to have greater access to the security forces than do elites in
personalist and single-party regimes. In military dictatorships, most members of the elite
coalition are military commanders of various forces. These commanders usually have
troops and weaponry at their personal disposal. As a result, the security forces tend to be
at least partially controlled by members of the elite coalition. Leaders have less control
over the security apparatus and individual commanders of various forces have more
autonomy.
This is not the case in single-party and personalist dictatorships. Elites in singleparty and personalist regimes do not have the same direct access to the security forces as
elites in military regimes do. In single-party regimes, the party makes great efforts to
keep the security apparatus under party control. They indoctrinate soldiers in the party
ideology and promote them according to party loyalty. Similarly, in personalist
dictatorships, the leader’s control over the security apparatus is often a central reason
why such regimes are identified as personalistic.10 Personalist dictators generally have
control over military promotions and can eliminate individuals they deem to be disloyal.
Though a member of the military may be part of the elite coalition in single-party and
personalist regimes, most elites do not directly control branches of the security forces.
This makes staging a coup more difficult in these regimes. Because elites in military
regimes have greater access to forces that can potentially unseat the leader, the
probability that military elites can successfully do so is greater.
Kebschull (1994) notes that when conditions are favorable, only a very small military force is actually
needed to achieve success.
10
In Geddes’s dataset (2003), for example, one of the criteria for labeling a regime personalistic is whether
the leader has personal control of the security police.
9
10
For these reasons, I argue that military elites have the greatest ability to oust the
dictator, followed by single-party elites, and lastly personalist elites. My argument is
summarized in Table 1.
[Insert Table 1 about here]
I have focused primarily on differences in the ability of the elite coalition to
overthrow the leader across dictatorships. Though this is a central element of elite-leader
relations in authoritarian regimes, it is by no means the only one. The ability to oust, for
example, is not synonymous with the desire to oust. Elites capable of unseating the
dictator may choose not to do so if they are satisfied with the dictator’s leadership.
Dictators must constantly act to ensure the continued support of elites in the coalition and
deter them from defecting to another “employer.” The strategies dictators pursue to curb
the desire of elites to unseat them are another important element of elite-leader relations.
My main focus in this study, however, is on the ability of the elite coalition to oust the
dictator.
III. Testing the Argument.
If the above reasoning is correct, then leadership ouster in dictatorships should be
strongly influenced by the institutional structure of the elite coalition. Elites in military
regimes should have the greatest ability to oust the dictator, followed by elites in singleparty regimes, and lastly elites in personalist regimes. This argument is tested
quantitatively by looking at the survival rates of dictators. I use Chiozza and Goemans’s
(2004) dataset, which includes entry and exit dates of all leaders holding executive power
for the 1919-1999 period. Leaders leaving office due to natural death are coded as
censored observations.11 Following Marinov (2005), I drop instances in which multiple
This approach is consistent with other work on the subject (see Marinov 2005, Bienen and van de Walle
1991, and Bueno de Mesquita et al. 2003.
11
11
leaders held office during a single calendar year. In these cases, I keep a single countryyear which records: 1) that a leadership change occurred, and 2) who the leader in power
is at the end of the calendar year.12 Formatting the data in this way is useful because data
on most covariates are measured annually.13
To classify authoritarian regimes, I rely on the codings of Geddes (2003). I
include in my dataset all dictatorships she lists as either personalist, military, or singleparty.14 Geddes includes regimes in existence since World War II that lasted three years
or longer, excluding monarchies. I create dummy variables for each regime type, with
single-party dictatorship the excluded dummy variable in all of the statistical tests. The
total number of regimes included in my dataset is 270, with 2358 observations. Total
sample size varies due to missing data.
The expectation is that leaders in military dictatorships should have the shortest
survival rates, followed by leaders in single-party dictatorships, and lastly leaders in
personalist dictatorships. Summary statistics of the longevity rates of authoritarian
leaders, presented in Table 2, support this expectation.
[Insert Table 2 about here]
As the table indicates, on average military leaders rule for 3.5 years, single-party leaders
rule for 8.8 years, and personalist leaders rule for 9.9 years. The following histograms
provide a visual distribution of the data.
[Insert histograms here]
12
I should note that predicting leadership turnover is somewhat distinct from predicting coup d'états.
While coups are generally clear indications of leadership turnover, leadership turnovers can occur without
violence. Leaders may voluntarily step down from power in order to avoid a violent overthrow. The threat
of a coup may be enough to induce a leader to leave office. For this reason, my dependent variable is
leadership removal, rather than coups.
13
Londregan and Poole (1996) use a similar approach.
14
Geddes also includes hybrid regimes in her dataset. I chose to exclude hybrid cases in order to provide a
more direct test of my argument. For more information regarding Geddes’s criteria, refer to Geddes 2003.
12
When dictatorships collapse, however, the leader of the regime often goes down
with it. To ensure the emphasis is on leadership failures, as opposed to regime failures, I
also conduct tests excluding the tenures of dictators whose removal from office coincides
with the demise of the regime. Summary statistics of dictators’ longevity rates excluding
these cases are presented in Table 3.15
[Insert Table 3 about here]
The basic result is virtually unchanged: on average, personalist dictators rule for about
three years longer than single-party dictators, and for about nine years longer than
military dictators.
A serious test, however, requires an array of economic and domestic political
controls to minimize the possibility that any correlation between regime type and
leadership removal is spurious. The first set of controls I use, GDP per capita (logged)
and economic growth, measure domestic economic conditions. Poor economic
conditions have been shown to negatively affect leadership survival (Londregan and
Poole 1990).16 Both of these variables are taken from the World Bank (2003).
Because various studies have shown that past coups lead to future coups
(Londregan and Poole 1990; Sanhueza 1999), in addition to the economic controls, I also
include a measure of the number of prior coups occurring in the last six years, as coded
by Chiozza and Goemans (2004). Bueno de Mesquita and Siverson (1995) find that
leaders who engage in war subject themselves to a domestic political hazard that
threatens their retention of political power. To account for this possibility, I control for
whether the country underwent a civil war, as measured by Fearon (2005), and whether
In this sample, leaders are excluded from the sample if they are removed from office during the last year
of the regime. Because leadership removal and regime collapse nearly always coincide in personalist
regimes, I do not exclude any personalist leaders.
16
This finding builds on the insight of Luttwak (1969) and Finer (1962) that economic backwardness is
essentially a necessary condition for coups.
15
13
the state was involved in a militarized interstate dispute in which force was applied
(force).17 Lastly, I include variables measuring the age of the leader, provided by
Chiozza and Goemans (2004), and the population of the country (logged), taken from
the Penn World Table (Heston, Summers, and Aten 2006), which have both been shown
to influence leadership survival (Marinov 2005; Chiozza and Goemans 2004). All control
variables are lagged one year to reduce problems of endogeneity.18
The estimation procedure I use is a Cox proportional hazard model. This is one of
the most general of the survival regression models because no assumptions are made
regarding the nature or shape of the hazard function. The Cox model estimates a “hazard
rate” for leadership removal at a particular moment.19 This hazard rate is modeled as a
function of the baseline hazard (h0) at time t, which is simply the hazard for an
observation when all independent variable values are equal to zero, and a number of
independent variables. Estimates of these covariates indicate proportional changes
relative to this baseline hazard. For example, a hazard ratio greater than 1 indicates an
increase in the risk of removal; a hazard ratio less than 1 indicates a decrease in the risk
of removal.
Though the Cox proportional hazard model is not based on any assumptions
regarding the shape of the underlying survival distribution, the model does assume a
multiplicative relationship between the underlying hazard function and the log-linear
function of the covariates (an assumption of proportionality). 20 In other words, it is
Source: Correlates of War data on Militarized Interstate Disputes (Jones, Bremer, and Singer 1996).
I also included in my models dummy variables measuring region, as some studies have shown that South
American countries are subject to a greater risk of coups than other countries (Londregan and Poole 1990;
Galetovic and Sanhueza 1995). Doing so did not alter the results. The results from these tests can be
obtained from the author upon request.
19
The hazard rate is defined as:
h(t) = probability of leadership removal between times t and t + 1
probability of leadership removal after time t
20
See Fox 2002.
17
18
14
assumed that the ratio of the hazard functions for two observations with different values
for independent variables does not depend on time. It is possible that this proportionality
assumption does not hold, however, and the impacts of some of my covariates depend on
time. Because of this possibility, I conduct an analysis of the scaled Schoenfeld residuals
for the primary models that I estimate. All of the covariates met the proportional hazard
assumption. For this reason, I do not include interactions between my explanatory
variables and the logarithm of time.21
In addition, I follow Chiozza and Goemans (2004, 607) and include a frailty term
in my specification. A frailty term is an additional unmeasured covariate αi that is
sampled from a Gamma distribution with mean 1 and variance θ that multiplicatively
affects the baseline hazard (see Therneau and Grambsch 2000, 231-260).22 It is similar to
a random effect, in that it assesses whether some leaders are more likely to be ousted, all
else equal. When αi is greater than 1, the leader is more likely to lose power than would
be expected, given the explanatory variables. When αi is smaller than 1, the opposite is
true. The frailty term essentially incorporates unmeasured heterogeneity across units. I
cluster observations by countries, so that the frailty term is constant within country
groups (Chiozza and Goemans 2004, 608). The idea is that leaders in some countries
may have systematically different frailties than leaders in other countries, independent of
the other covariates in the model.
Because some of the control variables I use may be predictors of the probability
of regime type (a potential source of selection bias), I present results from a Heckman
selection model in every table. This model controls for possible selection effects to
Graphs of the residuals and time also do not indicate non-proportionality.
The specification for the hazard takes the following form:
hi (t | x) = h 0i (t)αi exp (Σpj =1βj xj), where h0i (t) is the baseline hazard, and αi follows a Gamma distribution
with shape and scale parameters equal to 1/θ.
21
22
15
ensure that the regime type coefficients are not biased. To estimate the selectioncorrected effects of authoritarian regime type, I use the logged level of development,
civil war, dummy variables measuring region, and growth to generate multinomial logit
estimates of the authoritarian regime type, and then use this predicted result in my
models.23 I also include in each table models excluding each leader’s first year in office
and models excluding tenures in which regime and leadership failure coincide (as
discussed above). The results from these tests are presented in Table 4.
[Insert Table 4 about here]
Discussion of Results.
As the table indicates, personalist dictators have the lowest likelihood of being
overthrown and military dictators have the highest. According to Model 3, for example,
after correcting for selection effects, in comparison to single-party dictators, personalist
dictators are 64% less likely to be ousted and military dictators are 1000% more likely to
be ousted. These are substantively large differences. The effects are also statistically
significant at the .01 level. In short, these results support the theoretical framework that
I present.
It may be the case, however, that high leadership turnover is particularly likely
during the first year in office. To account for this, in Model 4 I exclude the first year of
the dictator’s tenure. The effects remain virtually unchanged. The same is true when
cases of simultaneous regime and leadership failure are excluded (Model 5).
The results also indicate that the number of prior coups in a country and the size
of the population increase a dictator’s risk of being ousted. In contrast, the age of the
leader leads to a lower hazard of being overthrown, as do economic growth and
Regions include Latin America, North Africa and the Middle East, Eastern Europe, Asia, and Africa.
Though fuel exports may also predict regime type (Ross 2001), due to problems of missing data, I do not
include such a measure in my selection model.
23
16
involvement in a military dispute (force). Interestingly, once selection effects are taken
into account, civil war and GDP per capita have almost no effect on a leader’s risk of
being deposed.
Robustness Checks.
In this section, I test whether my results are robust to an alternate classification of
institutions. I include in my statistical tests Bueno de Mesquita et al’s (BdM)
measurement of regime type, which focuses on the selectorate and the winning
coalition.24 BdM argue that institutions can be characterized by the size of the
selectorate (the set of people who have an institutional say in choosing leaders), denoted
by S, and the size of the winning coalition (the minimal set of people whose support the
incumbent needs in order to remain in power), denoted by W. BdM predict that it is
easier for leaders to survive when the size of W is small and the size of S is large. Table
5 presents results from tests in which measures of W and S are included in the
specification.
[Insert Table 5 about here]
As the table indicates, the effects of personalist dictatorship and military
dictatorship are still strong. According to Model 8, for example, after correcting for
selection effects, the hazard rate of a personalist dictator is the lowest and the hazard rate
of a military dictator is the highest. These differences are substantively large and the
effects are statistically significant. Interestingly, the effects of the BdM variables are also
in the direction expected, substantively large, and statistically significant. My findings
hold regardless of whether the first year of the leader’s tenure is excluded or cases of
simultaneous regime and leadership failure are excluded. These results are remarkably
robust and provide strong support for my argument. Even when controlling for other
24
See Bueno de Mesquita 2003 for information on the measurement of these variables.
17
measures of institutions, among dictatorships, military dictators face the highest risk of
being ousted, followed by single-party dictators, and lastly personalist dictators.
IV. Conclusion.
In this paper, I have examined a key element of authoritarian politics: leadership
survival. I argue that under dictatorships, one of the key political actors is the elite
coalition—the group that a dictator relies on for support. One of the key determinants of
the risk that dictators face on their hold on power is the nature of the elite coalition.
Whether elites are bound together by a dominant institution, like a party or military,
affects their ability to overthrow the dictator. I argue that there are two main factors that
affect the elite coalition’s ability to oust. The first is whether elites share membership in
a unifying institution. Membership in a dominant party or military, as in single-party and
military regimes, makes it easier for elites to coordinate to overthrow the leader. The
second factor is whether elites have control over the security forces, as in military
regimes. Direct influence over the security apparatus makes it easier to stage a coup.
The expectation follows that elites in military regimes should have the greatest ability to
oust the dictator, followed by elites in single-party regimes, and lastly elites in personalist
regimes.
I test this expectation quantitatively by using a survival model to predict the
ouster of dictators. Controlling for other relevant factors, I find strong support for my
argument. Among dictatorships, military leaders face a significantly greater risk of being
removed from power. On average, they only rule for around three or four years. The
average lifespan of single-party dictators is about double the average lifespan of military
dictators, but about a year shorter than that of personalist dictators. In fact, personalist
dictators face the lowest risk of being overthrown, ruling on average for around ten years.
18
These results are remarkably robust. They hold under a variety of circumstances, as well
as when alternate measures of domestic institutions are taken into account.
My findings are evidence in support of the claim that authoritarian institutional
structures matter and affect the relative power of the dictator vis-à-vis the elite coalition.
Dictatorial institutions play an important role and have significant consequences for these
regimes. When elites can credibly threaten to oust the dictator, this decreases the
bargaining power of the dictator relative to his supporters. This study is a small step
toward a deeper understanding of authoritarian politics.
Further research is necessary to
test differences in the ability of authoritarian support coalitions to make their will felt in
other policy areas.
19
Military Elites
Single-party Elites
Personalist Elites
Table 1: Power of Elite Coalition
Collective Actor?
Security
Ability to Stage Successful
Access?
Coup:
Yes
More likely
High
Yes
Less likely
Medium
No
Less likely
Low
20
Table 2: Survival Rates of Authoritarian Leaders, in years
Personalist Regimes
Single-party Regimes
Military Regimes
N
64
128
78
21
Mean
12.1
8.5
2.7
St Dev.
8.9
8.9
2.9
Median
11
5
2
Table 3: Survival Rates of Authoritarian Leaders, in years
( cases of simultaneous regime and
failure excluded )
N
Mean St Dev. Median
Personalist Regimes
64
12.1
8.9
11
Single-party Regimes
105
8.8
9.2
5
Military Regimes
46
2.7
3.0
2.7
22
Histograms
0
5
10
Percent
15
20
25
30
Lifespan of Military Leaders (in years)
0
5
10
15
20
lifespan
25
30
35
40
0
5
10
Percent
15
20
25
30
Lifespan of Single Party Leaders (in years)
0
5
10
15
20
lifespan
23
25
30
35
40
0
5
10
Percent
15
20
25
30
Lifespan of Personalist Leaders (in years)
0
5
10
15
20
lifespan
25
30
35
40
Table 4.
Leadership
Removal in
Authoritarian
Regimes
Model 1
Model 2
Model 3
(selection
corrected)
Model 4
(selection
corrected, first
year of tenure
excluded)
Personalist
dictatorship
Military
dictatorship
Civil war
.52*
(.19)
24.8***
(11.2)
.45**
(.17)
20.3***
(9.0)
2.1**
(.76)
2.0***
(.27)
.95***
(.01)
.05**
(.07)
1.11
(.17)
1.16
(.15)
.61
(.21)
1699
78
144
-804.3
.36***
(.12)
11.0***
(4.7)
1.1
(.46)
1.9***
(.24)
.96***
(.01)
.01***
(.01)
.99
(.18)
1.2
(.16)
.59
(.21)
1699
78
144
-791.3
.38***
(.13)
11.1***
(5.0)
1.0
(.45)
1.8***
(.24)
.96***
(.01)
.009***
(.01)
.98
(.18)
1.1
(.16)
.62
(.22)
1526
77
134
-731.6
Prior coups
Age
Growth
GDP per capita
(logged)
Population
(logged)
Force
No. of obs.
No. of countries
No. of failures
Log-likelihood
2086
84
163
-965.8
24
Model 5
(selection
corrected, cases
of simultaneous
regime and
leadership failure
excluded)
.45**
(.18)
19.4***
(10.0)
1.2
(.55)
2.0***
(.28)
.96**
(.01)
.008***
(.01)
.94
(.18)
1.2
(.17)
.60
(.24)
1477
74
112
-601.8
Wald χ2
87.22
132.15
166.63
154.73
126.18
2
(p > χ )
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
Θ
1.95
1.36
.96
.98
.98
Likelihood ratio test of θ = 0
χ2
128.46
54.61
34.86
36.20
27.23
(p > χ2)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
Results of a Cox proportionate hazard model; hazard ratios (standard errors). *p<.10, **p<0.05, ***p<0.01, twotailed tests. The Wald test refers to a test of the hypothesis that all coefficients are simultaneously equal to 0. The
frailty parameter θ measures the variance of a Gamma distribution with mean equal to 1.
Table 5.
Leadership
Removal in
Authoritarian
Regimes—
Robustness check
Personalist
dictatorship
Military
dictatorship
Civil war
Prior coups
Age
Growth
GDP per capita
(logged)
Population
(logged)
Force
Winning Coalition
Selectorate
Model 6
Model 7
Model 8
(selection
corrected)
Model 9
(selection
corrected, first
year of tenure
excluded)
3.5***
(.01)
2.0***
(.24)
.97**
(.01)
.12
(.18)
1.86***
(.32)
1.31*
(.19)
.67
(.23)
5.3**
(4.0)
.06***
(.02)
.42**
(.16)
10.0***
(4.7)
2.5**
(.96)
2.1***
(.30)
.95***
(.01)
.06**
(.08)
1.27
(.20)
1.27
(.20)
.61
(.22)
4.9**
(3.6)
.08***
(.03)
.33***
(.11)
4.9***
(2.24)
1.57
(.66)
2.1***
(.27)
.96**
(.01)
.02**
(.03)
1.2
(.22)
1.31**
(.17)
.62
(.22)
4.4**
(3.1)
.09***
(.03)
.34***
(.12)
4.5***
(2.1)
1.68
(.74)
2.0***
(.28)
.96***
(.01)
.02*
(.05)
1.22
(.24)
1.25*
(.17)
.65
(.24)
6.7**
(5.1)
.06***
(.02)
25
Model 10
(selection
corrected, cases
of simultaneous
regime and
leadership
failure
excluded)
.29***
(.13)
5.7***
(3.5)
1.14
(.56)
2.1***
(.32)
.97*
(.01)
.01**
(.02)
1.1
(.23)
1.33**
(.19)
.72
(.29)
2.0
(1.9)
.07***
(.03)
No. of obs.
1666
1666
1666
1498
1444
No. of countries
77
77
77
76
73
No. of failures
140
140
140
130
108
Log-likelihood
-773.9
-748.2
-735.8
-689.4
-549.0
Wald χ2
130.0
166.70
205.6
154.2
159.3
(p > χ2)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
θ
1.74
1.38
.94
1.45
1.12
Likelihood ratio test of θ = 0
χ2
90.0
65.49
41.50
64.22
42.51
(p > χ2)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
Results of a Cox proportionate hazard model; hazard ratios (standard errors). *p<.10, **p<0.05, ***p<0.01,
two-tailed tests. The Wald test refers to a test of the hypothesis that all coefficients are simultaneously equal to
0. The frailty parameter θ measures the variance of a Gamma distribution with mean equal to 1.
References
Acemoglu, Daron and James Robinson. 2001. “A Theory of Political Transitions.”
American Economic Review (91).
Arceneaux, Craig. 2001. Bounded Missions: Military Regimes and Democratization in
the Southern Cone and Brazil. University Park: Penn State University Press.
Arendt, Hannah. 1951. The Origins of Totalitarianism. New York: Harvest Books.
Bertocchi, Graziella, and Michael Spagat. 2001. “The Politics of Co-optation.” Journal
of Comparative Economics 29 (4): 591-607.
Bienen, Henry. 1978. “Military Rule and Political Processes: Nigerian Examples.”
Comparative Politics, 10 (2): 205-225.
Bienen, Henry, and Nicolas Van de Walle. 1991. Of Time and Power: Leadership
Duration in the Modern World. Stanford, CA: Stanford University Press.
Boix, Carles. 2003. Democracy and Redistribution. Cambridge University Press: New
York.
Bratton, Michael and Nicolas Van de Walle. 1997. Democratic Experiments in Africa:
Regime Transitions in Comparative Perspective. Cambridge: Cambridge University
26
Press.
Brooker, Paul. 1995. Twentieth-Century Dictatorships: the Ideological One-Party
States.
Washington Square, N.Y.: New York University Press.
Brooker, Paul. 2000. Non-Democratic Regimes: Theory, Government and Politics.
London: MacMillan Publishing.
Bueno de Mesquita, Bruce and Randolph M. Siverson. 1995. “War and the Survival of
Political Leaders: A Comparative Study of Regime Types and Political
Accountability.” American Political Science Review 89 (4).
Bueno de Mesquita, Bruce, Alastair Smith, Randolph M. Siverson, and James D.
Morrow. 2003. The Logic of Political Survival. Massachusetts Institute of
Technology: MIT Press.
Chebabi, H.E., and Juan J. Linz. 1998. Sultanistic Regimes. Baltimore: The Johns
Hopkins University Press.
Chiozza, Giacomo, and Henk E. Goemans. 2004. “International Conflict and the Tenure
of Leaders, Is War Still Ex Post Inefficient?” American Journal of Political Science
48 (3): 604-619.
Chorley, Katharine Campbell. 1973. Armies and the Art of Revolution. Boston: Beacon
Press.
Clapham, Christopher and George Philip (eds.). 1985. The Political Dilemmas of
Military Regimes. London: Croom Helm.
Decalo, Samuel. 1976. Coups and Army Rule in Africa: Studies in Military Style. New
Haven: Yale University Press.
Farouk-Slugett, Marion and Peter Slugett. 1987. Iraq since 1958: From Revolution to
27
Dictatorship. London: I.B. Tauris.
Fearon, James. 2005. “Primary Commodity Exports and Civil War.” Journal of
Conflict
Resolution 49 (4), 483-507.
Finer, Samuel. 1975. “State and Nation-Building in Europe: The Role of the Military,”
in The Formation of Nation States in Western Europe, ed. Charles Tilly. Princeton,
NJ: Princeton University Press.
Fitzpatrick, Sheila. 1982. The Russian Revolution. New York: Oxford University Press.
Fox, John. 2002. “Cox Proportional-Hazards Regression for Survival Data.”
<http://cran.r-project.org/doc/contrib/Fox-Companion/appendix-cox-regression.pdf >
Friedrich, Carl, and Zbigniew Brzezinski. 1956. Totalitarian Dictatorship and
Autocracy. Cambridge: Harvard University Press.
Gallego, Maria, and Carolyn Pitchik. 2004. “An Economic Theory of Leadership
Turnover.” Journal of Public Economics 88 (12): 2361-2382.
Galetovic, A. and Ricardo Sanhueza. 1995. “Citizens, Autocrats, and Plotters: An
Agency Theory of Coup d'état.” Working Paper, DII Universidad de Chile and
Universidad de Los Andes.
Gandhi, Jennifer. 2003. “Dictatorial Institutions and their Impact on Economic
Performance.” Presented at the American Political Science Association Annual
Meeting. Philadelphia, PA, August 28-31, 2003.
Geddes, Barbara. 2003. Paradigms and Sand Castles: Theory Building and Research
Design in Comparative Politics. Ann Arbor: University of Michigan Press.
Geddes, Barbara. 2004. “Minimum-Winning Coalitions and Personalization in
Authoritarian Regimes.” Presented at the American Political Science Association
28
Annual Meeting. Chicago, IL, September 2-5, 2004.
Grossman, Herschel I. and Suk Ja Noh. 1994. “Proprietary Public Finance and
Economic Welfare,” Journal of Public Economics 53 (2): February.
Haber, Stephen. 2006. “Authoritarian Government.” In Barry Weingast and Donald
Wittman (eds.), The Oxford Handbook of Political Economy. Oxford: Oxford
University Press.
Haggard, Stephen and Richard Kaufman. 1995. The Political Economy of Democratic
Transitions. Princeton, NJ: Princeton University Press.
Hartlyn, Jonathan. 1998. The Struggle for Democratic Politics in the Dominican
Republic. Chapel Hill: The University of North Carolina Press.
Higley, John, and Michael G. Burton. 1989. “The Elite Variable in Democratic
Transitions and Breakdowns.” American Sociological Review 54 (1): 17-32.
Heston, Alan, Robert Summers and Bettina Aten. 2006. Penn World Table Version 6.2,
Center for International Comparisons of Production, Income and Prices at the
University of Pennsylvania.
Huntington, Samuel P. 1970. Political Order in Changing Societies. New Haven: Yale
University Press.
Jones, Daniel M., Stuart A. Bremer and J. David Singer. 1996. "Militarized Interstate
Disputes, 1816-1992: Rationale, Coding Rules, and Empirical Patterns." Conflict
Management and Peace Science 15(2).
Langston, Joy. 2004. “When Do Ruling Parties Split? Evidence From Mexico in
Comparative Perspective.” Paper presented at the annual meeting of the American
Political Science Association, Chicago, IL, September 2, 2004.
Linz, Juan J. 1975. Totalitarianism and Authoritarianism. Boulder, CO: Lynne Rienner.
29
Londregan, John B. and Keith Poole. 1990. “Poverty, the Coup Trap, and the Seizure of
Executive Power.” World Politics, 42 (2 Jan): 151-183.
Londregan, John B. and Keith Poole. 1996. “Does High Income Promote Democracy?”
World Politics 49: 1-30.
Luttwak, Edward. 1969. Coup d’état; a Practical Handbook. New York: Knopf.
Kebschull, H.G. 1994. “Operation ‘Just Missed’: Lessons from Failed Coup Attempts.”
Armed Forces and Society 20, 4.
Kugler, Jack, and Yi Feng. 1999. “Explaining and Modeling Democratic Transitions.”
The Journal of Conflict Resolution 43 (2): 139-146.
Magaloni, Beatriz. 2006. Voting for Autocracy: Hegemonic Party Survival and its
Demise in Mexico. Cambridge: Cambridge University Press.
Marinov, Nikolay. 2005. “Do Economic Sanctions Destabilize Country Leaders?”
American Journal of Political Science, 94 (3).
McGuire, Martin C. and Mancur Olson, Jr. 1996. “The Economics of Autocracy and
Majority Rule: The Invisible Hand and the Use of Force.” Journal of Economic
Literature 34(1).
Nordlinger, Eric A. 1977. Soldiers in Politics: Military Coups and Governments.
Englewood Cliffs, NJ: Prentice-Hall Inc.
Olson, Mancur. 2000. Power and Prosperity: Outgrowing Communist and Capitalist
Dictatorships. Oxford University Press.
Pinckney, Robert. 1972. Ghana under Military Rule, 1966-1979. London: Methuen and
Co., Ltd.
Ross, Michael L. 2001. “Does Oil Hinder Democracy?” World Politics: April 2001.
Sanhueza, Ricardo. 1999. “The Hazard Rate of Political Regimes.” Public Choice,
30
March (98).
Stepan, Alfred. 1974. The Military in Politics: Changing Patterns in Brazil. Princeton
NJ: Princeton University Press
Svolik, Milan. 2006. “A Theory of Government Dynamics in Authoritarian Regimes.”
Unpublished manuscript.
Therneau, Terry M. and Patricia M. Grambsch. 2000. Modeling Survival Data:
Extending the Cox Model. New York: Springer-Verlag.
Tullock, Gordon. 1987. Autocracy. Boston: Kluwer Academic Publishers.
Wintrobe, Ronald. 1998. The Political Economy of Dictatorship. Cambridge:
Cambridge University Press.
World Bank. 2003. World Development Indicators 2003. Washington D.C.: World
Bank. CD-ROM.
Zolberg, Aristide. 1966. Creating Political Order: The Party-States of West Africa.
Chicago: Rand McNally.
31